The HD Camera Family

The original article on www.hd24.com

Like digital photography, HD requires a degree of discipline from all aspects of shooting including post production and choice and calibration of the display device.

When shooting HD with a compressed recording for a HD delivery you're on the edge. There are no safety ropes. So what is a HD picture? Is it defined by the type of camera or the recording format employed? Is there a way of knowing the providence of a HD image? It is possible to upconvert a picture from a cctv camera to HD. On playback, technically speaking you'll be watching a so called, HD picture. Lets examine the types of cameras so that you know in general where you are starting. Just how HD is your HD camera?

The pedigree of the picture starts with the lens.

A "proper" HD lens will deliver the shadow detail and overall resolution that is needed for a good HD picture. Be warned that some rental companies suggest that prime lenses designed for standard definition are acceptable for HD. Just like standard definition there are no standards for HD lenses and some are better than others, as a general rule the recent 2nd generation offerings are superior.

The Panavision and Fuji Super HD lenses are superb. They are fast T1.6 and designed to integrate seamlessly into a 35mm method of operation. However this his comes at a cost of size and weight when compared to other HD lenses.

To create a fast zoom lens for a 2/3 inch image plane the front element has to be large. Generally speaking the larger the front element the less depth of field. The 2/3 inch image size of HD is less than half that of 35mm motion picture film. As a result a wider lens is required on HD to match a given angle of view on 35mm.. Wider lenses have greater depth of field, this greater depth of field is both a help and a hindrance. Great for low light work. Not so good for controlling the sharpness of backgrounds in other situations. Depth of field for a given angle of view at T4 on 35mm compares to T1.6 on 2/3 inch chip

We have become accustomed to out of focus backgrounds in movies in modern times and associate sharper backgrounds with TV news and soaps. Interestingly sharp backgrounds were in vogue in the 40s, for example look at Gregg Toland's work. The introduction of less sensitive colour films stocks meant a change in practice, using lenses with wider apertures meant that backgrounds became softer.

Canon Fuji and Angenieux initially produced refined standard definition lenses to cater for the immediate demand of the Sony HDW700 HD news camcorders in 1997. Then came the Panavision lens for Star Wars and Fuji then introduced a set of primes that rivaled the resolution and functionality of the Panavision lenses. Canons second generation zooms launched in 2002 have been designed from the ground up with HD in mind.

Apart from the lower depth of field an audience will not notice the difference between Panavision, Ziess Fuji or Canon. Bluescreen work would benifit from

Fuji's New Super zooms, the Cook and the Angenieux Optimo are rivals in respect of speed to the Panavison Zoom lenses but they too are big and heavy. The front diameter of the Fuji Super zoom is 200 mm! and the Optimo weights 11 kilos. A full 35mm size crew should be employed when using these lenses.

The cost of these lenses is reflected in the rental rates which in some cases is more than the rental of the camera.

Zeiss entered the fray in 2002 with Digi Primes, considered the highest quality set of prime lenses at a fast aperture and launched a 6 to 24mm zoom in 2004.

In April 2003 Cooke optics presented a useful fast zoom. T1.7 8-46mm. Front element is 150mm and weight is 6 kilos. Unfortunatly a 6 inch matt box is needed with this lens.

The Panavision lenses are not anamorphic, so how did Lucas make a cinemascope aspect ratio image from a 16 x 9 frame? He "extracted" a Sony term, used to describe cropping the center portion of the frame. He started with 1080 lines in camera he ended up with 700 lines to transfer to film.

It is not surprising then, that today a good crew, shooting academy aspect ratio and working with latest lenses can surpass the horizontal resolution of Star Wars.

In May 2004 Canon launched a converter that fitrs between HD lens and camera that refocuses the cinemascope aspect ratio of the image then squashes it to fill the 16 x 9 frame. The other option for cinemascope is to use a Viper in scope mode, where full height 1920x1080 pixels are reassigned to 2.35:1 aspect ratio.

Some productions chose HD because they have a high shooting ratio, but in some cases a large and heavy zoom is counter productive as its mass makes it quite difficult to quickly reset the camera.

A production team should think carefully about the crew size , schedule and location before deciding which camera system or lenses to use.

The other way to create a narrow depth of field is to use an optical relay lens. These have been used in standard definition for years.

Basically a lens that is designed for a larger format say 35mm motion picture lenses is fixed to an adapter that refocuses the image onto the 2/3 inch ccds. The narrower depth of field of the larger format lens is maintained. The adapter absorbs 1 stop of light and usually degrades the image a little.

The latest version of this device features a fast spinning ground glass where the image is formed from a 35mm format lens. The image is then refocused onto the 2/3 inch chips. It creates a unique soft texture, once again with a loss of light.

After the lens, next in line is a prism. This splits the light into three separate image planes. There doesn't seem to be much difference between the prisms used by manufacturers. Fixed to the prism with incredible accuracy are three ccd sensors. The ccd has been the choice of sensor for the last 12 years but is now being challenged by CMOS sensors. However Fuji have recently announced some remarkable breakthroughs in ccd design for 35mm digital still cameras, including a ccd that has two pixels per photo site. One is large the other small. The small one naturally is less sensitive to light (by 2 stops) the outputs of the two photodiodes are combined by processing to produce a increase in dynamic range to rival film, according to Fuji anyway! It is feasible that this technology will find its way into video cameras.

The ccd is packed with millions of photo diodes. To form a PAL standard definition frame there are around 750Hx 625V diodes on each of 3 ccds forming a picture, in a HD camera there are either 1920Hx1080V or 1280H x 720V pixels on each of the ccds. Except for the Viper that is which has a total of 9 million pixels more on the Viper later.

There are studio cameras that are multi standard but there are no camcorders that are multi standard ie that record both 720p and 1080p formats.

All of the camcorders have removable smart media that store camera set-ups and use the B4 mount (except Panavised f900) Some of the box cameras use c mount lenses.

All can use a range of batteries and camera accessories such as LCD monitors. Panasonic use the Sony style base plate, a version of which has been redesigned by Chrozial so it is very solid indeed. Matt boxes are a nightmare if you mix and match base plates as the base plates vary in thickness.

The advent of the Panavision Sony Genesis opens a new door and is the first digital cinematography camera to crawl from the video swamp and not be clubbed to death as it hit the beach.

The Sony HDWf900 cine alta camera, HDW750 and HDW730 camcorders record on the HDCAM format. This format records a 1440Hx1080V frame. How do Sony convert 1920H to 1440H? They are very coy about this "sampling" it is described as a kind of coil compression.

As the human eye is more attuned to vertical than horizontal resolution this loss of quality is less than you would imagine. On replay the frame is converted to 1920x1080.

How much horizontal resolution is lost on a HD recording? Try monitoring a live HD camera on a 24 inch monitor, recording a scene and switching between live and playback.

Most of the time the viewer cannot see a difference between the live output and the playback. The live output from the camera head is full quality 1920 x1080. However there are artifacts with both the sampling and the HDCAM compression that can be seen by a computer trying to key a blue screen. Usually in the form of a single pixel between blue screen and subject that represents.... neither.

Fast motion against blue screen should be shot at a higher shutter speed to reduce motion blur, then motion blur added after the key is made in post.

The HDW 750 and HDW730 series has an inbuilt 1920x1080 HDSDI output as well as a down convertor (SDI) output whereas the f900 needs adapters. The f900 has 12 bit processing and the 750/730 is 10 bit. The f900 has a greater range of tones and colours at the expense of a larger heavier and more power hungry camera not to mention it is more expensive, certainly if you add up the cost of down conversion and HDSDI output adapters.

The 750p and 730 have a loop record function as an optional extra, where a RAM buffer holds 8 seconds of video and audio as in the DVW790. So it is possible to roll tape thus recording the previous 8 seconds, a useful feature for wildlife and news cameramen.

This feature is also used for timelapse but exposure duration does not exceed 1/25th second..

The HDSDI output via a single cable carries full quality uncompressed HD picture, with timecode and audio in perfect synch.

The f900 has a analogue HD output that needs 3 cables to carry the picture and separate cables for sound and timecode. The upside is that this Y PB PR signal requires no additional expensive decoding card in the monitors. But one BNC cable is defiantly easier to work with than three.

The f900 has undergone various upgrades the latest is a significant improvement to dynamic range and a lower noise level. Producers should check exactly what upgrades have been installed if quality is a issue. The mark 3 camera has a different block that eliminates green flare that some times arises when shooting pinpoints of light. A cine gamma curve is available for mark 3 cameras.

Panavised f900 The Panavised f900 is a modified f900, including a new front and bottom plate which is more secure than the standard Sony plate. The camera is designed to take the weight of the Panavision lenses.

The PV mount is fitted which necessitates the removal of white balance, shutter, roll buttons and menue wheel to a separate box.

A custom viewfinder and a new handle, designed with steadycam in mind are employed, along with a new high resolution colour viewfinder made by Accuscene.

A 35mm crew will feel at home with the Panavison camera lenses and accessories although some of the versatility of using video is traded for this design approach. The Panavison advertising is glossy and at times it is easy for the novice to get the wrong impression for instance claims that Panavision cameras achieve a wider colour gamut are hard to prove and many non video types think that SWAOTC was shot with anamorphic lenses. Panavision are fitting B4 lens mounts to some of their lenses so they can be used on HDW750ps in Europe for TV drama. It is uncertain if they will make these B4 mount lenses available to other users.

The new Accuscene colour viewfinder will initially be available exclusively to Panavison.

The Panasonic AJ-HDC 27Varicam has three 1 million pixel IT chips to form a 1220 x720 image at up to 60fps onto the Panasonic DVCPROHD format. The camera head employs 10 bit processing, 12 bit was not achievable in a small package due to heat and power considerations.

DVCPROHD varicam codec samples the image from 1220 x 720 to 960 x 720.

The difference between 10bit and 12bit is more obvious on a large screen or with carefully crafted lighting on smaller HD screens. Skin tones in particular look less plastic with 12 bit A to D processing.

The Varicam cannot produce a "native" 1920x 1080 HD signal from the camera head as it does not have the 2 million pixel ccds required. It does have ramping, time-lapse and native 60 fps recording. In timelapse mode a 1/4 of a second exposure is possible. Not quite long enough for taillights of speeding cars to join up and become a continuos snake of light but better than the Sony 750 and 730.

Numerous studio camera heads can output 60fps but the Varicam is the only camcorder with this feature.

The menue is well thought out and offers a range of features cinematographers have been asking for for years. The gamma curve is the closest yet to mimicking the response of film. That is not to say ccd cameras can have the same response as film! A gamma curve does not change the inherent characteristics of the ccd!

There is a useful real time readout of stops under and over that is useful to DPs who are new to video. The DVCPROHD system is lower cost than HDCAM because the recorder can only record 60fps. However the ccd can be scanned from 4 to 60 fps. All the time however the recorder is still humming along at 60fps. So if the DP wants 4 frames per second fine but the recorder needs 60fps to work so frames are repeated to pad the 4fps up to 60fps.

At the same time it puts a flag in the timecode part of the recording so in the edit only the key frames are extracted and the padded frames are ignored. If one is recording 60 fps then no frames are thrown away, if one is recording 10 frames per second then 50 frames are thrown away.

DVCPRO tapes are cheaper and smaller than HDCAM tapes.

The camera is small and light and is more sensitive due to its larger sized pixels. The 1920x1080 cameras need 2.1 million pixels where as the 720p cameras need only 1 million so it follows that more light reaches each pixel of a 720 camera. On the other hand use of gain on a higher res HD camera is a little less noticeable on a down converted picture than if originated on a 720p or standard def camera.

The Varicam would have taken off with more of a bang had its lower price been supported by a more straight forward post route.

The Viper is the cine style variant of the Thomson Grass Valley LDK6000. It is designed with a film style of operation in mind but not sacrificing the operational advantages of video. It is not a camcorder.

It is a compact head with a ccz style connector on the rear.

It is the only camera at the moment to have a output direct from the ccds. The current terminology to describe this camera is that it is a camera with a 444 output.

Well what is 444? In TV land we have to operate gain, detail and matrix setting in camera to select a narrow band of information that fully occupies the somewhat limited bandwidth of sampled TV recordings.

When I first started using digital slomo cameras in the mid nineties I was told that I could select gain, detail and matrix after the recording! A revelation.

444 enables virtually all of the output of the ccds to be recorded,(if you have a hard disk recorder that can cope with the huge data stream) so gain, detail matrix and gamma settings can be adjusted in post production. This means that highlight control and colour grading is performed in post rather than in camera. Typically this results in a one stop increase in dynamic range and more scope to manipulate colours.

Although the camera outputs raw data from its ccds it is still restricted to a modest 1920x1080 pixel output with the ccds sampled at 12 bit and converted to 10 bit log.

On location the DP need only control exposure, focus, frame rate, shutter and framing.

Some DPs do not like the concept of their images being altered in post, for others post grading it is just another tool and an opportunity for a few more days work.

We could have had 444 cameras years ago but there was little to be gained as TV sets aren't high fidelity displays and there was no convenient way to record 444. With the advent of video to film transfers the prospect of actually needing all the information from the ccd started some head scratching. The advent of The Directors Friend in 2000 which was the first portable Hard disk HDSDI recorder designed for location use, began the ball rolling. It is the size of an electric piano and is connected by cable to the camera. Its immediate use was recording uncompressed HDSDI from any HD camera.

A year later The Directors Friend was adapted to record 444.

Like most HD cameras and camcorders the Viper also has a HDSDI output which can also be recorded either on a uncompressed disk system or a compressed tape format.

The Viper is also unique in that it can scan and output 1920x1080 in 2.35:1 aspect ratio. It can achieve this by using ccds consisting of 1920H x 4320V pixels. Wow a super HD camera! No... unfortunately for the moment at least, the camera can only output 1920x 1080 pixels or pixel groups at a time. It groups the vertical sub pixels together so that the vertical line count can be set to 720, 1080 within a 16 x 9 format or 1080 within a cinemascope format of 2.35:1

It is the only camera to have this feature of full height (1080) cinemascope. This output can be 444 or HDSDI. The sheer quality of the data is a draw back for long form work although Sony's new HDCAM SR format field recorder SRW1 will be a natural partner for the Viper and indeed Sony's and others 444 cameras.

In the meantime the Viper offers studio based HD cinemascope format with extraordinary depth of field under low light being recorded on the HD tape format of choice or on the digital piano if you can find a post house willing to have a go with dealing with the data.

In HDSDI mode it has full control of gammas, detail, matrixes ect. It has a useful zoom function on the viewfinder to check focus.

Sony HDC F950

This is a upgraded HDC950 camera with 444 output. There is a T option for this camera, that splits the ccd block from the camera making a small camera head. It does not have a 10bit log output like the Viper.

Other Cameras

Hitachi SK3300p Studio camera 1080i 720p.

Hitachi SK9100P Studio camera 1080i only.

Hitachi DK-H3A small 3 chip B4 mount box camera for graphics and rostrum 1080i.

Ikegami HDL-20 a tiny 2 chip remote head HD camera, initially 59.94 with a 24p version coming soon with great promise since it will use small c mount lenses.

Ikegami HDL10 single chip remote head 1035i HD camera.

Ikegami HDL40 3 chip CMOS box camera

Ikegami HDK-79EX camera 720p or 1080i only studio camera.

Panasonic AK-HC900 box camera with B4 mount producing native 720p.

Panasonic-HDC 20A camcorder 10 bit, 59.94 only.

Sony HDW700 HDCAM camcorder (discontinued) 10 bit 59.95 only 1035i

Sony HDW700A HDCAM camcorder (discontinued) 10 bit 59.95 only 1080i

Sony HDW750 discontinued 60i only camera, not to be confused with the 750p which is 25p /50i camera.

Sony HDW 730 replacement for 750 it records 50i as well as 60i. Now renamed the HDW730S

Sony HDC950 Sony's studio camera. A remote head kit, HKC-T950, can be bought to make this the smallest 1920 x 1080 camera.

Sony HDCf950 New version of HDC 950 to be released 2004 with 444 output will also work in 444. "T' version for remote head mode.

Sony DXC H10 1920 x 1035i One of the first HD box cameras.

Sony X300 3 chip HD box camera using 1/2 inch chips 1440 x1080 pixels.

Thomson LDK6000 Mark II multi format, 24p capable studio camera. A Viper without 444. A very popular OB and studio HD camera.

JVC DZ-VCA3U a 3 chip remote head mains only medical camera with c mount.

1125 lines 750 x 750 pixels. Marginal HD quality.

Dalsa Origin the first of the non TV standard digital high resolution cameras. It uses a Bayer chip to create colours on its single chip.

Arri D20 Single chip CMOS sensor 2k camera using Bayer filter.

Kinetta, offers the promise of intercgangeable sensor and dockable hard drive to record 444.

Genesis, Sony Panavision 12 million pixel single chip ccd camera in super 35mm format.

 

Mind your Ps

It is possible to deinterlace a 60i picture turning the 2 field per frame into a one field frame image. By definition this becomes a progressive frame, albeit it a progressive frame with half the resolution removed! By then doubling the remaining lines a new picture is made which is in resolution terms is softer than the original but produces smooth motion when replayed at 24 or 25fps.

Rate card The ATSC television system includes four basic digital image formats &endash; two high definition and two standard definition. The 1920x1080 and 1280x720 images are high definition while 704x480 and 640x480 are standard definition. These are the transmission standards that get the picture from broadcaster to home set. In respect of live broadcasts the truck needs to output the standard that is being transmitted. It is possible to convert a 1920x1080 camera signal in the truck to another format.

Some cameras have dual output HD and standard def. Within these formats there are varying frame rates.

Every manufacturer on the planet calls 720p a HD format. Everyone but Sony who don't have a way of recording it even though they do have a studio camera that is 720p switchable. Upconversion or uprezing from the many digital and HD formats in the edit to 1920x1080 is not uncommon. Conversion cards are available for ccus to convert 60i cameras to "P" formats. This is not the same as scanning the ccd in "p" mode, far from it!

So we must be very careful of the pedigree of the picture:

What quality the lens?

How many ccd chips?

Number of effective pixels on the chip?

What is the "native" scanning rate of the ccd?

Are they sampled?

What bit depth?

Recorded at what compression rate?

On which recording format?

What losses in the edit to get from camera format to editing format?

The numbers are one thing, looking at the picture is another and in most cases the more revealing . It goes without saying that you must know the specification of the display device that you are using to measure your standard. Use your eyes!

Accept that our eyes play tricks, we have poor colour memory we are more sensitive to tones in the skin than blue tones in the sky. We see some artifacts and miss others.

These factors are not at all represented in the numbers comparisons between cameras.

Since major manufacturers are keen not to make a distinction between 1440 x 1080 and 1220 x720 and 1920x 1080, they only have themselves to blame when a $4k "HD" consumer camera gets plenty of attention! I refer to the JVC HDV camera.

When it comes to evaluating HD, use your head and your eyes.

Michael Brennan